Interview with Frank van Harmelen in NRC on Responsible Data Science: An Algorithm to determine whether you are suitable for that job…
In brief (full article here in Dutch):
Imagine that every vacancy you click on will state the chances you have with your profile that you will be invited to a job interview. If you do not have a chance, it also suggests which training you could follow to increase your chances. Handy, right?
In 2018, this scenario will be possible for all candidates of YoungCapital employment agency….. “If you do not get a job, we’ll let you know immediately what other jobs you can expect more success with,” says Rogier Thewessen, one of the founders of the agency.
The technique that makes it possible is machine learning. Simply explained, it means that you have a lot of examples for a computer to recognize patterns. Recruiters are excited about it. At one glance you see which job applicants are most suitable…
ADS Senior Researcher Frank van Harmelen:
An algorithm is determining whether you can work somewhere or not is making influential decisions. These are different from web store algorithms that recommend products. Algorithms that make these kind of socially sensitive decisions are still widely discussed in the research world, says Frank van Harmelen, professor at the VU Amsterdam and specialised in artificial intelligence. “The most important question is how to deal with sensitive data in a responsible manner…”
Take something like prejudices. The algorithm makes predictions based on the recruiters’ decision-making behavior, van Harmelen explains. “There are many assumptions about who is the best candidate. But how such a final decision is made is not always clear at all. Suppose that recruiters unconsciously prefer a woman for certain functions, then the computer takes that idea…”
Nevertheless, it remains challenging for machine learning that is often no longer a matter of how predictions have been made, stresses van Harmelen. “There are millions of complex calculations behind which even the makers of an algorithm can not explain correctly. Whilst it is just as important for a job applicant to hear from a recruiter why they are not invited. ”